Saturday, January 24, 2004

Not content with all of the tracking polls (five by my count: CNN/USA Today, ARG, Suffolk U., Zogby/Reuters/MSNBC, and Boston Globe), the media is now beginning to look at the one day results of the tracking polls. Usually, a tracking poll surveys 200 people a day and then combines the results for three days. This gives a sample of about 600 people--a decent sample size and a margin of error (MOE) of about +/- 4 percentage points. A one-day sample of 200 people, however, gives a MOE of +/-7 percentages. That's a very wide margin, so wide that if Kerry were at 28, Dean 18, Clark 17, Edwards 15, and Lieberman 14, the correct headline would be "Five-Way Tie for New Hampshire Lead!."

To show how the one day results can jump around, Suffolk U. is giving the daily results of their tracking poll. Here are Friday's results:

Kerry 35
Dean 18
Clark 11

But on Thursday, it was:

Kerry 23
Dean 22
Clark 18

The point is, don't trust one day results.

No comments: